VR simulation data intelligent classification method

Abstract

The invention relates to the technical field of data processing, in particular to a VR simulation data intelligent classification method. According to the scheme, the multiple keywords in the historical VR simulation scene can be obtained, the corresponding features of the keywords are obtained so as to calculate the similarity between the keywords, the corresponding features and the membership degree between the keywords are obtained, the classification weight of each keyword is obtained by using the corresponding features and the membership degree between the keywords, and the classification efficiency of the keywords is improved. The method comprises the steps of obtaining classification weights of current simulation data, obtaining decision tree importance through the classification weights to train an XGboost model, and inputting the current simulation data into the trained XGboost model to realize classification of the current simulation data. According to the method, the final classification result of the simulation data in the current VR simulation scene is determined according to the similarity between the simulation scene keyword prepared in the current VR simulation scene and the keyword obtained by the XGboost model, so that the classification rate and accuracy of the VR simulation data are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method for intelligent classification of VR simulation data. Background Art

[0002] With the application of VR simulation technology in various industries, its types and forms are becoming increasingly diverse. VR simulation technology provides powerful tools for scientific research, education, entertainment, and other fields by constructing highly realistic virtual environments to simulate various scenarios and tasks in the real world. The construction of VR simulation environments usually involves multiple parallel tasks and complex interactive relationships, which require efficient and accurate data management to support them. Data classification and hierarchical management, as a currently recognized effective data management method, plays an important role in VR simulation environments. Through intelligent data classification, different tasks, scenarios, and objects can be accurately identified, enabling VR simulation systems to handle multiple parallel tasks and complex situations, thereby improving simulation efficiency and accuracy.

[0003] VR simulation tasks typically involve multiple simulation scenarios, each with distinct data requirements. For example, urban traffic flow analysis involves simulation scenarios such as peak-hour traffic flow analysis, analysis of the impact of road construction on traffic flow, and analysis of traffic flow under new traffic control measures. These scenarios are closely interrelated and mutually influential, and each has unique data requirements. However, existing data classification methods often fail to fully consider the correlation between simulations in VR simulation scenarios and the degree of affiliation between VR simulation data and simulation scenarios. This neglect can lead to inaccurate classification of VR simulation data by classification models, thus affecting the reliability and validity of simulation results.

[0004] Therefore, there is an urgent need for a method that can effectively solve the problem that the existing technology does not fully consider the correlation between simulations of VR simulation scenes and the membership between VR simulation data and simulation scenes, which may lead to inaccurate classification of VR simulation data by the classification model. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent classification method for VR simulation data to solve the problem that the existing technology does not fully consider the correlation between simulations of VR simulation scenes and the membership between VR simulation data and simulation scenes, which may lead to inaccurate classification of VR simulation data by the classification model; to this end, the present invention provides the following technical solutions.

[0006] The present invention provides a method for intelligent classification of VR simulation data, comprising: Acquire multiple keywords in a historical VR simulation scene, obtain corresponding features for the keywords, calculate similarities between the keywords, and obtain membership between the corresponding features and the keywords, wherein the similarities between the keywords are obtained by calculating the similarities of feature histograms corresponding to the keywords, and the similarities of the feature histograms are obtained by statistically analyzing multiple features contained in the keywords, and the membership is used to characterize the degree of association between the keywords and the corresponding features; The classification weight of each keyword is obtained by using the membership between the corresponding features and keywords, and the importance of the decision tree is obtained by the classification weight to train the XGboost model; The current simulation data is input into the trained XGboost model to classify the current simulation data.

[0007] In the above scheme, the present invention obtains multiple keywords in the historical VR simulation scene, obtains corresponding features for the keywords, calculates the similarity between the keywords, obtains the membership between the corresponding features and the keywords, uses the membership between the corresponding features and the keywords to obtain the classification weight of each keyword, and uses the classification weight to obtain the decision tree importance to train the XGboost model, uses the membership between the corresponding features and the keywords to obtain the classification weight of each keyword, and uses the classification weight to obtain the decision tree importance to train the XGboost model, so as to realize the classification of the current simulation data, and effectively improve the classification rate and accuracy of the current simulation data.

[0008] Optionally, the calculation process of the membership degree is: ; in, is the membership of the f-th feature to the i-th keyword in the k-th simulation scene; is the similarity between the feature histograms of the i-th keyword and the j-th keyword containing f features in the k-th simulation scene; H k is the total number of keyword combinations in the kth simulation scenario, is the mean value function, is the number of other features under the ith keyword after removing the fth feature in the kth simulation scene. The cosine similarity between the feature and the feature under the j-th keyword.

[0009] In the above scheme, the degree of association between the keywords and the corresponding features is obtained by calculating the membership degree.

[0010] Optionally, the similarity between the feature histograms specifically includes: ; in, For the The similarity of the feature histograms corresponding to the i-th keyword and the j-th keyword in the simulation scene; and Respectively The feature histogram corresponding to the i-th keyword and the j-th keyword in the simulation scene text; c is the ... j-th keyword in the simulation scene text; The cth feature in the feature histogram with the largest feature statistics among the i-th keyword and the j-th keyword in the simulation scene; and Respectively The usage ratio of the cth feature in the feature histogram corresponding to the i-th keyword and the j-th keyword in the simulation scene.

[0011] The above scheme calculates the similarity between keywords by calculating the similarity of the feature histograms corresponding to the keywords. Therefore, the similarity between keywords can intuitively reflect the similarities and differences between keywords. Using the similarity of the feature histograms, we can more accurately determine the degree of association between the corresponding features and keywords.

[0012] Optionally, the classification weight of each keyword is calculated as follows: ; in, is the classification weight of the i-th keyword of the historical simulation data corresponding to the f-th feature in the historical VR simulation scene; is the membership of the f-th feature to the i-th keyword in the k-th simulation scene; n is the total number of keywords in the k-th simulation scene; K is the total number of historical VR simulation scenes.

[0013] The above solution improves the accuracy of XGboost model decision making by calculating the classification weights of the keywords.

[0014] Optionally, the decision tree importance is obtained in the following manner: ; in, is the decision importance of the vth decision tree in the XGboost model; is the classification accuracy of the historical simulation data assigned to the v-th decision tree.

[0015] The above scheme can optimize the accuracy of the XGboost model for classifying the current simulation data by calculating the importance of the decision tree.

[0016] Optionally, the classification accuracy of the historical simulation data is obtained by: ; in, is the classification accuracy of the historical simulation data assigned to the vth decision tree; is the total number of keywords of the fth feature; is the total number of types of historical simulation data assigned to the vth decision tree; is the classification weight of the i-th keyword of the historical simulation data corresponding to the f-th feature in the historical VR simulation scene; The classification accuracy of the historical simulation data corresponding to the f-th feature on the v-th decision tree is classified as the i-th keyword.

[0017] Optionally, the keywords are obtained by using Jieba tools to segment the text of the historical VR simulation scene.

[0018] Optionally, the corresponding features are features obtained after standardization and normalization preprocessing of data of different dimensions in the simulation scene.

[0019] Optionally, the XGboost model is a model established using VR simulation data.

[0020] Optionally, after the corresponding features are preprocessed, spline interpolation is used to complete the missing parts of the corresponding features.

[0021] The beneficial effects of the present invention are: The solution of the present invention can obtain multiple keywords in historical VR simulation scenes, obtain corresponding features for the keywords, calculate the similarity between the keywords, obtain the membership between the corresponding features and the keywords, use the membership between the corresponding features and the keywords to obtain the classification weight of each keyword, and use the classification weight to obtain the decision tree importance to train the XGboost model, and input the current simulation data into the trained XGboost model to classify the current simulation data. The present invention determines the final classification result of the simulation data in the current VR simulation scene based on the similarity between the simulation scene keywords proposed in the current VR simulation scene and the keywords obtained by the XGboost model, effectively improving the classification rate and accuracy of VR simulation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart schematically illustrates the steps of a method for intelligent classification of VR simulation data in this embodiment. DETAILED DESCRIPTION

[0023] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0024] It should be understood that the order of execution of the steps in the following embodiments does not necessarily mean the order in which they are executed. The order in which each process is executed should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0025] The present invention is directed to a data classification management solution for a VR simulation environment. The VR simulation data intelligent classification solution improves the accuracy of VR simulation data classification by calculating the similarity between keywords and obtaining the membership between the corresponding features and keywords.

[0026] Therefore, based on the above content, the present invention provides a VR simulation data intelligent classification method to improve the accuracy of VR simulation data classification.

[0027] Specifically, if Figure 1 As shown, a VR simulation data intelligent classification method in this embodiment includes the following steps: Step S1, obtaining multiple keywords in a historical VR simulation scene.

[0028] Specifically, VR simulation tasks usually have multiple simulation scenarios, and each scenario has different data requirements. For example, if the VR simulation scenario is urban traffic flow analysis, the simulation scenarios involved may include: traffic flow analysis during peak hours, the impact of road construction on traffic flow, and traffic flow under new traffic control measures. Therefore, it is necessary to extract keywords from the description text of the historical VR simulation scene and pre-process the corresponding features of the obtained keywords, wherein the pre-processing includes standardizing and normalizing the corresponding features of the obtained keywords. After pre-processing, if there are missing values ​​for the corresponding features, spline interpolation is used to fill them in to facilitate the extraction of the corresponding features.

[0029] The corresponding features include the geometric and physical features required in the simulation scene (features that describe the shape, size, position of objects in the virtual environment and the physical behavior of simulated objects in the virtual environment), perceptual features (including corresponding features such as texture, color, lighting, shadow and sound effects in the environment), interaction features (relevant features of the input device collecting the user's operating instructions and interaction intentions) and situational and plot features (features that describe the background, atmosphere and storyline of the virtual environment and define information such as the sequence of events, character behavior and task objectives in the virtual environment). In other words, the corresponding features in the present invention are different depending on the simulation scene created. The keywords refer to the key text words or phrases used when describing or constructing the simulation scene. These keywords play an important role in clarifying scene features, guiding scene construction and subsequent scene applications.

[0030] For example, the text of the simulation scene describes the traffic network of the city in a geographic VR simulation scene, and focuses on analyzing the traffic flow during peak hours. The scene description includes: the dynamic changes of busy streets, main roads, intersections and traffic lights. Users can use VR devices to observe the impact of traffic density on the road, vehicle flow speed and changes in traffic signals on the overall flow during the simulated peak hours; as well as environmental characteristics, weather conditions (sunny, rainy, etc.), audio features, such as traffic signal sounds, alarm sounds, etc., and image features, such as urban spatial structure, building structure, etc. By obtaining the description of the corresponding features in the simulation text, keywords are obtained based on the description. The examples described in the present invention are for illustrative purposes only and do not constitute any limitation to the present invention.

[0031] In this embodiment, by obtaining the corresponding features and keywords required in historical VR simulation scenes, the core elements and key information of the VR simulation scenes can be quickly understood, thereby avoiding blind searches in large amounts of data, and based on historical corresponding features, new VR simulation scenes can be designed more accurately to ensure that task scenarios, parameter settings, etc. are closer to actual needs, thereby improving the practicality and effectiveness of tasks.

[0032] Step S2: Obtain corresponding features for the keywords to calculate the similarity between the keywords and obtain the membership between the corresponding features and the keywords.

[0033] Specifically, a VR simulation scene text may contain multiple keywords. Therefore, it is necessary to calculate the similarity between the keywords to obtain the membership between the corresponding features and the keywords. The similarity between the keywords is obtained by calculating the similarity of the feature histograms corresponding to the keywords. The similarity of the feature histograms is obtained by statistically analyzing the various features contained in the keywords. The membership is used to represent the degree of association between the keywords and the corresponding features.

[0034] Step S21: Calculate the similarity between feature histograms. The specific calculation process is as follows: ; in, For the The similarity of the feature histograms corresponding to the i-th keyword and the j-th keyword in the simulation scene; and Respectively The feature histogram corresponding to the i-th keyword and the j-th keyword in the simulation scene text; c is the ... j-th keyword in the simulation scene text; The cth feature in the feature histogram with the largest feature statistics among the i-th keyword and the j-th keyword in the simulation scene; and Respectively The usage ratio of the cth feature in the feature histogram corresponding to the i-th keyword and the j-th keyword in the simulation scene.

[0035] Step S22: Calculate the degree of membership between the corresponding feature and the keyword based on the similarity of the feature histogram.

[0036] Specifically, when the similarity of the feature histogram is obtained, that is, the similarity of the keyword, the membership between the corresponding feature and the keyword is calculated: ; in, is the membership of the f-th feature to the i-th keyword in the k-th simulation scene; is the similarity between the feature histograms of the i-th keyword and the j-th keyword containing f features in the k-th simulation scene; H k is the total number of keyword combinations in the kth simulation scenario, is the mean value function, is the number of other features under the ith keyword after removing the fth feature in the kth simulation scene. The cosine similarity between the feature and the feature under the j-th keyword.

[0037] It is the mean similarity between the feature histograms of the i-th keyword and the j-th keyword containing f kinds of features in the k-th simulation scenario. The larger the value, the smaller the difference in required features between the i-th keyword and the j-th keyword containing the f-th corresponding feature, the greater the correlation degree of the f-th corresponding feature under the i-th keyword, and the smaller the membership degree of the i-th keyword corresponding to the f-th corresponding feature. The smaller the value, the greater the difference in required features between the i-th keyword and the j-th keyword containing the f-th corresponding feature, the smaller the correlation degree of the f-th corresponding feature under the i-th keyword, and the greater the membership degree of the i-th keyword corresponding to the f-th corresponding feature.

[0038] is the cosine similarity between the feature type corresponding to the i-th keyword and the feature type corresponding to the j-th keyword in the k-th simulation scene after removing the f-th corresponding feature. is the mean cosine similarity between the feature type corresponding to the ith keyword and the feature types corresponding to the rest of the keywords after removing the fth feature in the k simulation scenarios. The larger the value, the more similar the feature type corresponding to the ith keyword and the feature types corresponding to the rest of the keywords after removing the fth feature. The greater the downward correction degree, the smaller the correlation degree of the fth corresponding feature under the ith keyword, and the greater the membership degree of the VR simulation data corresponding to the fth corresponding feature to the ith keyword. The smaller the value, the less similar the corresponding feature type under the ith keyword after removing the fth corresponding feature is to the feature types under the remaining keywords. The smaller the downward revision.

[0039] In this embodiment, the present invention obtains corresponding features for keywords to calculate the similarity between keywords and obtain the membership between the corresponding features and the keywords. The present invention calculates the similarity of the feature histograms corresponding to the keywords and can intuitively reflect the similarities and differences between the data by statistically analyzing the distribution of the features corresponding to the keywords. When calculating the membership between the features corresponding to the keywords and the keywords, the similarity of the feature histograms is used as a basis to more accurately judge the degree of association between the corresponding features and the keywords.

[0040] Step S3: Using the membership between the corresponding features and keywords, the classification weight of each keyword is obtained, and the decision tree importance is obtained through the classification weight to train the XGboost model.

[0041] Specifically, the formula for calculating the classification weight of each keyword using the membership degree is: ; in, is the classification weight of the i-th keyword of the historical simulation data corresponding to the f-th feature in the historical VR simulation scene; is the membership of the f-th feature to the i-th keyword in the k-th simulation scene; n is the total number of keywords in the k-th simulation scene; K is the total number of historical VR simulation scenes.

[0042] The importance of the decision tree is obtained by the classification weight, and its calculation formula is: ; in, is the decision importance of the vth decision tree in the XGboost model; is the classification accuracy of the historical simulation data assigned to the v-th decision tree.

[0043] The classification accuracy of the historical simulation data is obtained as follows: ; in, is the classification accuracy of the historical simulation data assigned to the vth decision tree; is the total number of keywords of the fth feature; is the total number of types of historical simulation data assigned to the vth decision tree; is the classification weight of the i-th keyword of the historical simulation data corresponding to the f-th feature in the historical VR simulation scene; The classification accuracy of the historical simulation data corresponding to the f-th feature on the v-th decision tree is classified as the i-th keyword.

[0044] In this embodiment, the classification weight of each keyword is obtained by utilizing the membership between the corresponding features and keywords, and the decision importance of the decision tree is obtained by the classification weight. The importance of the decision tree is then used to train the XGboost model to improve the accuracy of the XGboost model in classifying simulation data.

[0045] Step S4: using the decision importance of the decision tree to train an XGboost model, and inputting the current simulation data into the trained XGboost model to classify the current simulation data.

[0046] In this embodiment, the XGboost model is trained using the decision importance of the decision tree, and the current simulation data is input into the trained XGboost model so that the trained XGboost model classifies the VR simulation data. The decision importance extracted by the decision tree can determine which features have the greatest impact on the classification results, thereby improving the accuracy of classification, and reducing unnecessary features, so that the XGboost model can become more concise, thereby improving the computational efficiency of the XGboost model.

[0047] The training process of the XGboost model is as follows: each type of VR simulation data and the corresponding keyword are input into the XGboost model so that the XGboost model predicts the classification results of the VR simulation data. The XGboost model is trained using a loss function, a cross entropy loss, and a gradient amplitude. When the XGboost model reaches the set maximum number of training times or the loss of the model is less than a preset loss threshold, the XGboost model stops training.

Claims

1. A VR simulation data intelligent classification method, characterized in that: include: Acquire multiple keywords in a historical VR simulation scene, obtain corresponding features for the keywords, calculate similarities between the keywords, and obtain membership between the corresponding features and the keywords, wherein the similarities between the keywords are obtained by calculating the similarities of feature histograms corresponding to the keywords, and the similarities of the feature histograms are obtained by statistically analyzing multiple features contained in the keywords, and the membership is used to characterize the degree of association between the keywords and the corresponding features; The classification weight of each keyword is obtained by using the membership between the corresponding features and keywords, and the importance of the decision tree is obtained by the classification weight to train the XGboost model; The current simulation data is input into the trained XGboost model to classify the current simulation data.

2. A VR simulation data intelligent classification method according to claim 1, characterized in that: The calculation process of the membership degree is: ； in, is the membership of the f-th feature to the i-th keyword in the k-th simulation scene; is the similarity between the feature histograms of the i-th keyword and the j-th keyword containing f features in the k-th simulation scene; H k is the total number of keyword combinations in the kth simulation scenario, is the mean value function, is the number of other features under the ith keyword after removing the fth feature in the kth simulation scene. The cosine similarity between the feature and the feature under the j-th keyword.

3. A VR simulation data intelligent classification method according to claim 2, characterized in that: The similarity between the feature histograms specifically includes: ； in, For the The similarity of the feature histograms corresponding to the i-th keyword and the j-th keyword in the simulation scene; and Respectively The feature histogram corresponding to the i-th keyword and the j-th keyword in the simulation scene text; c is the ... j-th keyword in the simulation scene text; The cth feature in the feature histogram with the largest feature statistics among the i-th keyword and the j-th keyword in the simulation scene; and Respectively The usage ratio of the cth feature in the feature histogram corresponding to the i-th keyword and the j-th keyword in the simulation scene.

4. The method for intelligent classification of VR simulation data according to claim 1, characterized in that: The calculation method of the classification weight of each keyword is specifically as follows: ； in, is the classification weight of the i-th keyword of the historical simulation data corresponding to the f-th feature in the historical VR simulation scene; is the membership of the f-th feature to the i-th keyword in the k-th simulation scene; n is the total number of keywords in the k-th simulation scene; K is the total number of historical VR simulation scenes.

5. The method for intelligent classification of VR simulation data according to claim 1, characterized in that: The method for obtaining the importance of the decision tree is specifically as follows: ； in, is the decision importance of the vth decision tree in the XGboost model; is the classification accuracy of the historical simulation data assigned to the v-th decision tree.

6. A VR simulation data intelligent classification method according to claim 4, characterized in that: The classification accuracy of the historical simulation data is obtained as follows: ； in, is the classification accuracy of the historical simulation data assigned to the vth decision tree; is the total number of keywords of the fth feature; is the total number of types of historical simulation data assigned to the vth decision tree; is the classification weight of the i-th keyword of the historical simulation data corresponding to the f-th feature in the historical VR simulation scene; The classification accuracy of the historical simulation data corresponding to the f-th feature on the v-th decision tree is classified as the i-th keyword.

7. The method for intelligent classification of VR simulation data according to claim 1, characterized in that: The keywords are obtained by segmenting the text of the historical VR simulation scene using the Jieba tool.

8. The method for intelligent classification of VR simulation data according to claim 1, characterized in that: The corresponding features are features obtained by standardizing and normalizing data of different dimensions in the simulation scene.

9. The method for intelligent classification of VR simulation data according to claim 1, characterized in that: The XGboost model is a model established using VR simulation data.

10. The method for intelligent classification of VR simulation data according to claim 8, characterized in that: After the corresponding features are preprocessed, the missing parts of the corresponding features are supplemented using spline interpolation.

Images